Ear recognition based on Multi-bags-of-features histogram

This paper proposes a novel image feature representation method, called multi-BOF histogram, for ear recognition. Given an ear image, we at first convolve it with J Gabor filters sharing the same parameters except the parameter of orientation. Then they obtained responses of each pixel at each scale and orientation can get J features. Then, each pixel can be assigned a unique features vector, namely “multi scale Gabor features vector)”. The classification is based on the image´s histogram. Extensive experiments conducted on the Delhi-I database demonstrate the overall superiority of Multi-bags-of-words histogram representation (M-BoF) over the other state-of-the-art ear representation methods evaluated.